Open Bug 1120607 Opened 10 years ago Updated 2 months ago

Consider changing the SurfaceCache expiration timer

Categories

(Core :: Graphics: ImageLib, defect, P3)

defect

Tracking

()

People

(Reporter: erahm, Unassigned)

References

(Blocks 1 open bug)

Details

(Whiteboard: [MemShrink:P2][awsy][gfx-noted])

As noted in bug 1104622, comment 10 there was a regression in the "RSS after TP5" metrics on 11/28 related to 1104622 landing. It's possible this is acceptable and we just need to annotate the regression on AWSY.
Assigning to seth based on the aforementioned bug 1104622 comment 10 to get this off the "incoming gfx bugs" radar.
Assignee: nobody → seth
Whiteboard: [MemShrink] → [MemShrink][awsy]
Whiteboard: [MemShrink][awsy] → [MemShrink:P1][awsy]
Seth recommends setting a custom pref [1] for AWSY so that we can work around this added volatility. I'll see if this helps us in the meantime. We should probably keep this open as a reminder to actually tweak that value properly when we get more stats. [1] "image.mem.surfacecache.min_expiration_ms"
Specifically, the stats we're waiting for are telemetry data about SurfaceCache usage. I'll file a tracking bug about that soon and will block this bug.
Any progress with the telemetry data, Seth?
Flags: needinfo?(seth)
(In reply to Nicholas Nethercote [:njn] from comment #4) > Any progress with the telemetry data, Seth? Nope. This is still on my radar but I haven't had a chance to move forward on it. I'll add it to my TODO list and try to at least *add* the telemetry relatively soon, so that when we're ready to act on it, the data we need is available.
Flags: needinfo?(seth)
Per discussion with Nick, retitling to make it more clear why this bug is still open, and dropping to P2. Basically what we want to know is: what's the optimal value of the "image.mem.surfacecache.min_expiration_ms" pref? The regression happened because we increased that value to 60 seconds, but obviously that means that we also hit in the cache more often, which benefits users. So what we want to do here is add some telemetry to try to make a more principled choice for this pref's value. FWIW, my guess is that the sweet spot would be to actually increase it further, but we need numbers.
Summary: 15MB - 25MB regression after bug 1104622 → Consider changing the SurfaceCache expiration timer
Whiteboard: [MemShrink:P1][awsy] → [MemShrink:P2][awsy]
Whiteboard: [MemShrink:P2][awsy] → [MemShrink:P2][awsy][gfx-noted]

The bug assignee didn't login in Bugzilla in the last 7 months, so the assignee is being reset.

Assignee: seth.bugzilla → nobody
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.